纹理特征提取的不变矩探讨  

Study on the moment invariants of texture feature extraction

作　　者：冯登超[1] 陈刚[1] 吴新颖[1] 杜文雅[2] 肖楷乐[1] 韩睿[1] 

机构地区：[1]北华航天工业学院电子工程系,廊坊065000 [2]太原科技大学电子信息学院,太原030024

出　　处：《国外电子测量技术》2012年第4期57-59,共3页Foreign Electronic Measurement Technology

基　　金：河北省自然科学基金项目(F2011409008);河北省教育厅青年基金项目(2010208);北华航天工业学院博士基金项目(KY-2008-02-B)联合资助

摘　　要：针对目标形状发生变化引起图像特征提取困难等问题,采用不变矩作为特征量的特征匹配算法,以纹理图像为研究对象,分析不变矩的基本理论,探讨具有全局统计信息的不变矩设计方法。仿真实验中,分别利用Hu矩和Zernike矩提取纹理图像的视觉特征,比较2种统计矩在纹理特征提取中的应用范围。实验结果表明,不变矩可以作为目标形状改变后的一种图像特征。然而,对于新的不变矩的构建仍需作进一步研究。Aimed at the difficult extraction of image feature caused by the shape change,invariant moments as the features of texture image was adopted in the paper to realize feature matching,where the basic theory of moment invariant feature matching algorithm was analyzed and the moment invariants method of global statistical information was studied.Texture image visual features was extracted by Hu moment and Zernike moment in the simulation experiment and two statistical moments in the texture feature extraction in the scope of application were compared as well.The experiment results show that the invariant moments can be used as the image feature with shape change.However,the new construction of invariant moments still needs further research.

关 键 词：纹理 特征匹配 HU矩 ZERNIKE矩 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]